goodboy
Your agent has no skill for the task? goodboy finds the best one in the top-starred GitHub skill repos, ranks it with a small trained model (no LLM, no API key), and loads only that skill.
Free
Works with the AI Agents you already use
goodboy
Example session with this skill installed
Review my Python code for security vulnerabilities. Is there a skill for that?
- Read your context and instructions
- Compiled the goodboy
Summary
I didn't have a security-review skill, so I used goodboy to find one. It searched GitHub, checked the top-starred skill repos, and found
Cloudflare's security-audit skill (70.6k stars) as a confident match. I've loaded it and I'm using it for your review.
What goodboy ran
$ goodboy fetch "review my python code for security vulnerabilities"
Searched GitHub for: python security vulnerabilities skill
Top 1 repos by stars that have a matching skill:
70.6k stars cloudflare/security-audit-skill (1 skill)
Fetched: cloudflare/security-audit-skill:security-audit (confidence 100%)
Saved to: ~/.local/share/goodboy/skills/cloudflare__security-audit-skill/security-audit
The skill it found
| Skill | security-audit |
| Repo | cloudflare/security-audit-skill |
| Stars | 70.6k |
| Confidence | 100% |
| Scripts included | none, so nothing will run without your OK |
What the skill covers: security guidance and vulnerability review for codebases, APIs, services, CLI tools, libraries and daemons. Its approach is to find vulnerabilities that cross a real trust boundary, then give the source evidence, a safe reproduction, a priority and the smallest effective fix.
Two modes, per the skill:
- Guidance mode (default): focused review of the code you point me at, with no files created.
- Full audit mode: all six phases plus written report artifacts, only if you ask for a full or end-to-end audit.
Next steps
- Tell me which files or folders to review, or say "full audit" for the complete six-phase workflow with a report.
- I'll go through them using the skill's method and list each finding with evidence, severity and a fix.
- For more options,
goodboy search "<task>"shows the other candidates it ranked.
Connects securely to your tools. The creator never sees your data.
About this skill
What it does
When your agent hits a task it has no skill for (a design system, a security audit, a slide deck, PDF work), goodboy goes and finds one. It searches GitHub live for the most-starred repos that actually ship SKILL.md skills on that topic, ranks every candidate, and downloads only the winning skill's folder for your agent to follow.
How it picks
- Search. Your task is cut down to topic keywords and searched on GitHub, sorted by stars.
- Verify. Each repo is checked for real
SKILL.mdfiles with a tree-only clone (file names only), so a 300-skill repo costs seconds, not a full download. - Rank. A learning-to-rank model trained for this one job scores every skill against your request. In testing on skills it never saw, it puts the right one first 90% of the time and in the top 3 97% of the time.
- Choose. The 3 most-starred repos with a matching skill are considered (the best match always keeps a slot), and only the chosen skill is downloaded.
- Report. Your agent tells you which skill it picked, from which repo, how many stars it has, and how confident goodboy is. If nothing fits, goodboy says so instead of guessing.
Why use it
- No LLM, no API key, no GPU. The ranker is about 2 KB of weights running in numpy; cached lookups take about half a second.
- Safe by default. goodboy only copies files and never runs anything. Fetched skills are treated as third-party content: the skill tells your agent to ask before installing, read scripts before running them, and ignore any fetched instruction to reveal secrets or exceed what you asked.
- Open source (MIT): github.com/Mamdouhelzein/goodboy
Requirements
Python 3.9+ and git. Your agent installs it once with pipx install git+https://github.com/Mamdouhelzein/goodboy, after asking you.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Requires Python 3.9+ and git. Unauthenticated GitHub searches are limited to ~10 per minute. Does not execute scripts; only fetches and reads skill files.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 5 days ago
- Free with an Agensi account
- Software
Python 3.9+
Git
Permissions
Terminal: runs the goodboy CLI. Network: GitHub search and git clones. Read/Write files: saves fetched skills under ~/.local/share/goodboy and reads them back. Environment: reads an optional GITHUB_TOKEN. goodboy never executes fetched skills or their scripts.